Robust Matsui observer tuning under parameter variations based on 2D experimental designs

This paper proposes an original solution for the tuning of a Matsui position observer on a permanent magnet synchronous sensorless drive. In such a redundant observer based on back emf estimation, there is no analytical way for coefficient tuning. The specific structure of this observer includes gains and filters that have to be tuned, in order to reach the best possible convergence. The tuning method in this paper is based on the analysis of parameter influences on a given performance criteria. The most influent parameters and their optimal level are identified and the optimal and robust values for gains and bandwidths are found under different load and velocity conditions with the help of an original 2D experimental design approach. Optimal sets of parameters are identified under large load variations and current measurement noises. As a result, the final sets of parameters achieve effective stability and satisfying convergence performance. First experimental results are shown on an experimental test bench.